U.S. patent application number 12/293062 was filed with the patent office on 2009-12-24 for wide-band equalization system.
This patent application is currently assigned to HARMAN INTERNATIONAL INDUSTRIES, INCORPORATED. Invention is credited to Ashish Aggarwal, Ulrich Horbach, Pedro Manrique.
Application Number | 20090316930 12/293062 |
Document ID | / |
Family ID | 38510280 |
Filed Date | 2009-12-24 |
United States Patent
Application |
20090316930 |
Kind Code |
A1 |
Horbach; Ulrich ; et
al. |
December 24, 2009 |
WIDE-BAND EQUALIZATION SYSTEM
Abstract
A Wide-band Equalization System ("WBES") based on near- and
far-field measurement data. The WBES includes a subwoofer equalizer
having an FIR filter together with decimator and interpolator
filters for processing low frequency signals. The WBES may also
include satellite channels for processing mid- and high-frequency
signals, where each satellite channel includes cascaded IIR filters
that process mid-frequency and high-frequency signals,
respectively. The WBES may also include a DSP that performs the
functions required by the IIR and FIR filters.
Inventors: |
Horbach; Ulrich; (Canyon
Country, CA) ; Aggarwal; Ashish; (Santa Clarita,
CA) ; Manrique; Pedro; (Pasadena, CA) |
Correspondence
Address: |
THE ECLIPSE GROUP LLP
10605 BALBOA BLVD., SUITE 300
GRANADA HILLS
CA
91344
US
|
Assignee: |
HARMAN INTERNATIONAL INDUSTRIES,
INCORPORATED
Northridge
CA
|
Family ID: |
38510280 |
Appl. No.: |
12/293062 |
Filed: |
March 14, 2007 |
PCT Filed: |
March 14, 2007 |
PCT NO: |
PCT/US2007/064006 |
371 Date: |
March 5, 2009 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
60782369 |
Mar 14, 2006 |
|
|
|
Current U.S.
Class: |
381/99 |
Current CPC
Class: |
H04S 2420/07 20130101;
H04S 7/301 20130101; H04S 7/307 20130101 |
Class at
Publication: |
381/99 |
International
Class: |
H03G 5/00 20060101
H03G005/00 |
Claims
1. A method for equalizing an audio system using near- and
far-field measurement data, the method comprising: capturing a set
of room impulse responses ("RIRs") at a plurality of listening
locations of the audio system; determining low-frequency finite
impulse response ("FIR") coefficients for a low-frequency FIR
filter; determining mid-frequency FIR coefficients for a
mid-frequency FIR filter; determining high-frequency FIR
coefficients for a high-frequency FIR filter; generating the
low-frequency FIR filter utilizing the low-frequency FIR
coefficients; generating the mid-frequency FIR filter utilizing the
mid-frequency FIR coefficients; generating the high-frequency FIR
filter utilizing the high-frequency FIR coefficients; generating an
at least one low-frequency filter of the audio system utilizing a
subwoofer equalizer ("EQ") that includes the low-frequency FIR
filter; generating an at least one mid-frequency filter of the
audio system as a plurality of cascaded infinite impulse response
("IIR") filters that are derived from the mid-frequency FIR filter;
and generating an at least one high-frequency filter of the audio
system as a plurality of cascaded IIR filters that are derived from
the high-frequency FIR filter.
2. The method of claim 1, wherein the subwoofer EQ further includes
a decimator filter and an interpolator filter.
3. The method of claim 1, wherein generating the low-frequency FIR
filter includes: determining a low-frequency inverse spectrum from
the captured set of RIRs; and multiplying the captured
low-frequency inverse spectrum by a target function that results in
an EQ filter frequency response.
4. The method of claim 3, wherein the target function is a bandpass
filter with 4.sup.th order low-pass and high-pass Butterworth
filter characteristics.
5. The method of claim 3, wherein determining the low-frequency
inverse spectrum further includes smoothing peaks of the EQ filter
frequency response utilizing a smoothing factor.
6. The method of claim 1, wherein generating the high-frequency FIR
filter coefficients includes: multiplying a near-field RIR derived
from the captured set of RIRs by a first time window; determining
the magnitude spectrum of the windowed near-field RIR; smoothing
the magnitude spectrum with a first smoothing factor; determining a
log-magnitude inverse spectrum of the smoothed magnitude spectrum;
smoothing the peaks of the log-magnitude inverse spectrum with a
second smoothing factor to derive a high-frequency EQ filter
spectrum; scaling the high-frequency EQ filter spectrum to a gain
equal to zero decibels at an operating frequency fg; limiting the
response of the high-frequency EQ filter spectrum to an upper
operating frequency fgu; clipping the gain of the high-frequency EQ
filter spectrum to a maximum allowed gain; determining an EQ FIR
filter impulse response out of the log-magnitude inverse spectrum;
and applying a second time window to the EQ FIR filter impulse
response.
7. The method of claim 6, wherein determining the EQ FIR filter
impulse response out of the log-magnitude inverse spectrum is
implemented utilizing a Hilbert transform.
8. The method of claim 6, wherein the second smoothing factor is
greater than the first smoothing factor.
9. The method of claim 1, wherein generating the mid-frequency FIR
filter includes: multiplying a far-field RIR derived from the set
of captured RIRs by a first time window; determining a magnitude
spectrum of the windowed RIR utilizing an N-point fast Fourier
transform ("FFT"); smoothing the magnitude spectrum with a first
smoothing factor; determining a log-magnitude inverse spectrum of
the smoothed magnitude spectrum; and determining an EQ filter
frequency response out of the log-magnitude inverse spectrum
utilizing a target function.
10. The method of claim 1, wherein the equalization of the
low-frequency signals, the mid-frequency signals, and the
high-frequency signals is performed simultaneously.
11. A Wide-band Equalization System ("WBES") for equalizing an
audio system using near- and far-field measurement data, the WBES
comprising: a bass manager in signal communication with a signal
source; a subwoofer EQ in signal communication with the bass
manager, and configured to receive low-frequency signals from the
bass manager; and a plurality of satellite channels in signal
communication with the bass manager, and configured to receive mid-
and high-frequency signals from the bass manager.
12. The WBES of claim 11, wherein the subwoofer EQ includes a
decimator filter, the at least one low-frequency FIR filter, and an
interpolator filter.
13. The WBES of claim 11, wherein each of the plurality of
satellite channels includes an at least one mid-frequency IIR
filter and an at least one high-frequency IIR filter, where the at
least one mid-frequency IIR filter and the at least one
high-frequency IIR filter are generated from the at least one
mid-frequency FIR filter and the at least one high-frequency FIR
filter, respectively.
14. The WBES of claim 13, further including a plurality of cascaded
IIR filters that are generated from the at least one mid-frequency
FIR filter and the at least one high-frequency FIR filter,
respectively.
15. A Wide-band Equalization System ("WBES") for equalizing an
audio system using near- and far-field measurement data, the WBES
comprising: means for capturing a set of room impulse responses
("RIRs") at a plurality of listening locations of the audio system;
means for determining low-frequency finite impulse response ("FIR")
coefficients for a low-frequency FIR filter; means for determining
mid-frequency FIR coefficients for a mid-frequency FIR filter;
means for determining high-frequency FIR coefficients for a
high-frequency FIR filter; means for generating the low-frequency
FIR filter utilizing the low-frequency FIR coefficients; means for
generating the mid-frequency FIR filter utilizing the mid-frequency
FIR coefficients; means for generating the high-frequency FIR
filter utilizing the high-frequency FIR coefficients; means for
generating an at least one low-frequency filter of the audio system
utilizing a subwoofer equalizer ("EQ") that includes the
low-frequency FIR filter; means for generating an at least one
mid-frequency filter of the audio system as a plurality of cascaded
infinite impulse response ("IIR") filters that are derived from the
mid-frequency FIR filter; and means for generating an at least one
high-frequency filter of the audio system as a plurality of
cascaded IIR filters that are derived from the high-frequency FIR
filter.
16. The WBES of claim 15, wherein the means for generating the
low-frequency FIR filter includes: means for determining a
low-frequency inverse spectrum from the captured set of RIRs; means
for multiplying the captured low-frequency inverse spectrum by a
target function that results in an EQ filter frequency
response.
17. The WBES of claim 16, wherein the means for determining the
low-frequency inverse spectrum further includes means for smoothing
peaks of the EQ filter frequency response utilizing a smoothing
factor.
18. The WBES of claim 15, wherein the means for generating the
high-frequency FIR filter coefficients includes: means for
multiplying a near-field RIR derived from the captured set of RIRs
by a first time window; means for determining the magnitude
spectrum of the windowed near-field RIR; means for smoothing the
magnitude spectrum with a first smoothing factor; means for
determining a log-magnitude inverse spectrum of the smoothed
magnitude spectrum; means for smoothing the peaks of the
log-magnitude inverse spectrum with a second smoothing factor to
derive a high-frequency EQ filter spectrum; means for scaling the
high-frequency EQ filter spectrum to a gain equal to zero decibels
at an operating frequency fg; means for limiting the response of
the high-frequency EQ filter spectrum to an upper operating
frequency fgu; means for clipping the gain of the high-frequency EQ
filter spectrum to a maximum allowed gain; means for determining an
EQ FIR filter impulse response out of the log-magnitude inverse
spectrum; and means for applying a second time window to the EQ FIR
filter impulse response.
19. The WBES of claim 15, wherein the means for generating the
mid-frequency FIR filter includes: means for multiplying a
far-field RIR derived from the set of captured RIRs by a first time
window; means for determining a magnitude spectrum of the windowed
RIR utilizing an N-point fast Fourier transform ("FFT"); means for
smoothing the magnitude spectrum with a first smoothing factor;
means for determining a log-magnitude inverse spectrum of the
smoothed magnitude spectrum; and means for determining an EQ filter
frequency response out of the log-magnitude inverse spectrum
utilizing a target function.
20. The WBES of claim 15, wherein the means for determining the
low-frequency, the mid-frequency, and the high-frequency FIR
coefficients includes a digital signal processor ("DSP").
21. The WBES of claim 15, wherein the means for generating the at
least one low-frequency filter of the audio system includes a
DSP.
22. The WBES of claim 11, wherein the means for generating the at
least one mid-frequency filter of the audio system includes a
DSP.
23. The WBES of claim 11, wherein the means for generating the at
least one high-frequency filter of the audio system includes a DSP.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application claims the benefit of U.S. Provisional
Application Ser. No. 60/782,369 entitled "Wide Band Equalization in
Small Spaces," filed Mar. 14, 2006, which application is
incorporated herein, in its entirety, by this reference.
BACKGROUND OF THE INVENTION
[0002] 1. Field of the Invention
[0003] The invention is generally related to an equalization system
that improves the sound quality of an audio system in a listening
room. In particular, the invention relates to an equalization
system that improves the sound quality of an audio system based
upon near- and far-field measurement data.
[0004] 2. Related Art
[0005] The aim of a high-quality audio system is to faithfully
reproduce a recorded acoustic event, such as a concert hall
experience, in smaller enclosed spaces, such as a listening room, a
home theater or entertainment center, a PC environment, or an
automobile.
[0006] The perceived sound quality of an audio system in smaller
enclosed spaces depends on several factors: quality and radiation
characteristics of the loudspeakers (e.g., on- and off-axis
frequency responses); placement of the loudspeakers at their
connect positions according to the standard (for example, ITU
5.1/7.1); acoustics of the room in general (low frequency modes,
reverb time, frequency-dependent absorption, effects of room
geometry and dimensions, location of furniture, etc.); and nearby
reflective surfaces and obstacles (e.g., on-wall mounting,
bookshelves, TV sets, etc.).
[0007] In order to provide an optimum listening experience in such
enclosed spaces, a digital "room equalization" system may be used.
In general, equalization is the process of either boosting or
attenuating certain frequency components in a signal. There are
several types of equalization, each with a different pattern of
attenuation or boost. Examples are a high-pass filter, bandpass
filter, graphic equalizer, and parametric equalizer.
[0008] In a multiband parametric equalizer ("EQ"), center
frequency, bandwidth (Q-factor) or peak shape, and gain (peak
amplitude above a given reference) in each of the bands may be
adjusted to flatten a measured frequency response at a listening
location (e.g., a seat in a listening room), Typically, a cascade
of second-order IIR ("infinite impulse response") filter sections
("biquads") is used to control frequency response. A digital signal
processor ("DSP") may generate test signals for each loudspeaker
(e.g., either white or pink noise or logarithmic sweeps), in order
to capture room responses at a desired listening location. For that
purpose, an omni-directional microphone may be positioned at the
listening location and connected to a signal analyzer or back to
the DSP.
[0009] In FIG. 1, a test system 100 that uses an equalizer to
produce a signal at the listening location that resembles the input
signal is shown. In an example of operation, signal source 104
produces a test signal, which is amplified by the preamplifier 106
and processed by the equalizer 108. The test signal is then
amplified by the power amplifier 110 and transmitted to a
loudspeaker 112. The loudspeaker 112 reproduces the test signal as
an acoustic pressure wave that is emitted from the loudspeaker 112,
which is then picked up by the test microphone 116 and passed to a
signal analyzer 120.
[0010] In this example of operation, the received test signal is
observed at the signal analyzer 120 and, in response, the test
signal may be adjusted accordingly through the equalizer 108. In
other implementations, the test microphone 116 may be directly in
signal communication with the equalizer 108, where the received
test signal may be automatically processed by the equalizer 108,
which may include digital signal processors ("DSPs"). Additionally,
the test microphone 116 may be positioned at a listening location
in a room or hall, where it can then capture the impulse responses
at that particular listening location.
[0011] In this example, if the equalizer 108 is a parametric EQ
with multiple filters, the multiple filters may be set manually, so
that, for example, a displayed response curve, on an output device
(not shown) in signal communication with the equalizer 108, becomes
smoother, or automatically, with the aid of an external processor
such as, for example a personal computer ("PC") or design logic
built into the DSP itself. In general, it is difficult and
suboptimal to adjust a set of cascaded parametric filter sections
because of overlap. Two or more of the parametric filter sections
may affect the same frequency band of interest, which leads to the
difficulty that a large number of parameters need to be adjusted
simultaneously. At low frequencies, it is important to accurately
suppress individual room modes. In order to avoid approximation
errors and quantization noise, a FIR ("finite impulse response")
filter may be directly used and operated at a low sample rate (for
example, utilizing decimation) to minimize processing cost.
[0012] In adjusting a frequency response, it is important to
distinguish between resonances (e.g., loudspeaker cabinet material
resonances, or standing waves at low frequencies in rooms) and
interferences due to multiple reflections that lead to nulls (dips)
in the frequency response. Resonances and room modes need to be
suppressed, e.g., with a notch filter, while narrow-band
interference dips strongly depend on the measurement position and
generally should be left unaltered. An attempt to correct
narrow-band interference dips may introduce high-gain peak filters
that are perceived as resonances.
[0013] In an intermediate frequency band (between approximately 100
Hz to 1000 Hz), it is desirable to correct errors related to the
source only, not the whole listening room. For example, eliminating
sonic differences between the main stereo speakers and the center
speaker, which may be close to a reflective surface such as a TV
set, leads to an improved stereo image. This so-called
"source-related" correction is independent of a particular
listening location, whereas a complete room correction would be
valid at a single point only.
[0014] At high frequencies (i.e., greater than 1000 Hz), the
in-room response is normally not flat, but decreases with
frequency. This may be addressed by a so-called "target function."
Equalization is performed such that the final response approximates
the target function. However, the correct target function choice
depends on the absorption properties of the particular room and the
radiation characteristics of the loudspeakers, and is thus a priori
unknown. In a (domestic) listening room solution, a set of
near-field measurements close to the loudspeakers provides
frequency response data above typically 1000 Hz, thus eliminating
the need for a target function. In all automobile, an adjustable
target function may be provided with the EQ algorithm.
[0015] Along with the foregoing considerations, there are many
other factors to be considered when trying to optimize the sound
quality audio systems utilized in small spaces such as listening
rooms or cars. Therefore, there is always a continuing need to
improve the sound quality of these audio systems, in particular, by
improving the fully-automated equalization of the responses of
loudspeakers located in these small spaces.
SUMMARY
[0016] A Wide-band Equalization System ("WBES") for equalizing an
audio system based on near- and far-field measurement data is
disclosed. The WBES may include a subwoofer EQ having an FIR filter
together with decimator and interpolator filters for processing low
frequency signals. The WBES may also include satellite channels for
processing mid- and high-frequency signals, where each satellite
channel includes cascaded IIR filters that process mid-frequency
and high-frequency signals. The WBES may also include one or more
DSPs that perform the functions required by the IIR and FIR filters
and may also generate test signals for a device under test.
[0017] In an example operation, the WBES may perform a method
whereby low-frequency, mid-frequency, and high-frequency FIRs are
generated from a captured set of room impulse responses ("RIRs"),
with a low-frequency filter of the audio system then implemented
using the low-frequency FIR, a decimator filter, and an
interpolator filter. Mid- and high-frequency filters of the audio
system may be implemented utilizing cascaded infinite impulse
response ("IIR") filters derived from the mid- and high-frequency
FIRs.
[0018] Other systems, methods, features and advantages of the
invention will be or will become apparent to one with skill in the
art upon examination of the following figures and detailed
description. It is intended that all such additional systems,
methods, features and advantages be included within this
description, be within the scope of the invention, and be protected
by the accompanying claims.
BRIEF DESCRIPTION OF THE FIGURES
[0019] The invention can be better understood with reference to the
following figures. The components in the figures are not
necessarily to scale, emphasis instead being placed upon
illustrating the principles of the invention. Moreover, in the
figures, like reference numerals designate corresponding parts
throughout the different views,
[0020] FIG. 1 shows a block diagram illustrating all example of a
known room equalization system.
[0021] FIG. 2 shows a block diagram illustrating an example of an
implementation of a Wide-band Equalization System ("WBES") in
accordance with the invention.
[0022] FIG. 3 shows a flow diagram illustrating an example of a
method performed by the WBES of FIG. 2 for correcting the response
of an individual loudspeaker based upon near-field, high-frequency
measurements.
[0023] FIG. 4 shows a graphical representation of an example of a
plot of amplitude versus time (in samples) of a raw (i.e.,
unwindowed) and a windowed impulse response produced by the method
described in FIG. 3.
[0024] FIG. 5 shows a graphical representation of an example of a
plot of the frequency response obtained using an N-point FFT
(N=8192), and the frequency response smoothed by a smoothing factor
produced by the method described in FIG. 3.
[0025] FIG. 6 shows a graphical representation of an example of
plots of the frequency responses of an ideal EQ filter, a smoothed
version of that frequency response, and the smoothed version with
those parts of the frequency response of an ideal EQ filter that
lie above the smoothed version of the frequency response cut from
the plot produced by the method described in FIG. 3.
[0026] FIG. 7 shows a graphical representation of an example of a
plot of a frequency response of an EQ filter impulse response that
has been scaled, limited to an upper frequency, and clipped to a
maximum gain by setting filter values above a defined gain value to
that value produced by the method described in FIG. 3.
[0027] FIG. 8 shows a graphical representation of an example of a
plot of amplitude versus time (in samples) of an EQ filter impulse
response that is time-limited produced by the method described in
FIG. 3.
[0028] FIG. 9 shows a graphical representation of an example of a
plot of frequency responses of an approximated IIR EQ filter
impulse response produced by the method described in FIG. 3.
[0029] FIG. 10 shows a graphical representation of an example of a
plot of frequency responses of a captured room impulse response, an
EQ filter impulse response, and the result of applying the EQ
filter impulse response to the captured room impulse response
produced by the method described in FIG. 3.
[0030] FIG. 11 shows a flow diagram illustrating an example of a
method performed by the WBES of FIG. 2 for correcting the response
of an individual loudspeaker based upon far-field, low-frequency
measurements.
[0031] FIG. 12 shows a graphical representation of an example of a
plot of amplitude versus frequency (in Hz) of an approximated
low-frequency FIR EQ filter impulse response produced by the method
described in FIG. 11.
[0032] FIG. 13 shows a flow diagram illustrating an example of a
method performed by the WBES of FIG. 2 for correcting the response
of an individual loudspeaker based upon far-field, mid-frequency
measurements.
[0033] FIG. 14 shows a graphical representation of an example of a
plot of amplitude versus time (in samples) of a windowed far-field
room impulse response produced by the method described in FIG.
13.
[0034] FIG. 15 shows a graphical representation of an example of
plots of amplitude versus frequency (in Hz) of a raw, measured and
a smoothed far-field spectrum at mid frequencies produced by the
method described in FIG. 13.
[0035] FIG. 16 shows a graphical representation of an example of
plots of amplitude versus frequency (in Hz) of a smoothed spectrum
and an EQ filter frequency response produced by the method
described in FIG. 13.
[0036] FIG. 17 shows a graphical representation of another example
of plots of amplitude versus frequency (in Hz) of low- and
mid-frequency EQ filter frequency responses produced by the method
described in FIG. 13.
[0037] FIG. 18 shows a graphical representation of an example of
plots of amplitude versus frequency (in Hz) of EQ filter frequency
response and room responses before and after room correction
produced by the method described in FIG. 13.
[0038] FIG. 19 shows a graphical representation of an example of a
plot of a frequency response of a target function produced by the
method described in FIG. 13.
[0039] FIG. 20 shows a graphical representation of an example of a
plot of the frequency responses of three bands of an EQ filter
produced by the method described in FIG. 13.
DETAILED DESCRIPTION
[0040] In the following description of examples of implementations
of the present invention, reference is made to the accompanying
drawings that form a part hereof, and which show, by way of
illustration, specific implementations of the invention that may be
utilized. Other implementations may be utilized and structural
changes may be made without departing from the scope of the present
invention.
[0041] In FIG. 2, a block diagram illustrating an example of an
implementation of a wide band equalization system ("WBES") 200 in
accordance with the invention is shown. WBES 200 may include
several signal processing modules that process low-, mid-, and
high-frequency signals. As an example of operation, a low frequency
signal 204 is generated by the bass manager 202, which may also
generate m mid- and high-frequency signals 206, where m typically
may be 5-7. The low-frequency signal 204 may be processed by a
subwoofer EQ 208 utilizing a room equalization algorithm. The
subwoofer EQ 208 includes a decimation filter 210, a subwoofer
equalizer FIR filter 212 of order n.sub.fir (typically
n.sub.fir=256 . . . 512), and an interpolation filter 214 to
resample the signal to the original sample rate (typically, the
decimation/interpolation ratio r=32 . . . 64).
[0042] Mid- and high-frequency signals 206 generated by the bass
manager 202 may be processed by "satellite" channels 216 1, 2, . .
. , and m (typically, m=5 or 7). Each satellite channel 216 may
include a cascade of mid-frequency-EQ second-order IIR biquad
sections 218 A_1, . . . , A_n.sub.1, and high-frequency-EQ biquads
220 sections B_1, . . . , B_n.sub.2, where, as an example,
n.sub.1=n.sub.2=3.
[0043] The filter coefficients for the mid-frequency-EQ IIR filters
218 and the high-frequency-EQ IIR filters 220 are based on measured
room responses and may be obtained by utilizing a room equalization
method. These IIR filters are higher order filters approximated
from mid- and high-frequency FIRs designed from far-field and
near-field measurement data. FIGS. 3, 11, and 13 illustrate
examples of room equalization methods used to obtain the filter
coefficients for the IIR and FIR filters shown in FIG. 2. These
room equalization methods may be implemented in a common DSP that
also performs real-time signal processing (i.e., the actual
filtering). Turning to FIG. 3, a flow chart illustrating an example
of a room equalization method is shown, where the room equalization
method is designed for a near-field, high-frequency EQ configured
to correct the impulse response of an individual loudspeaker and
its immediate surroundings in a room above approximately 1 kHz. The
process 300 starts in step 302 and in step 304, a room impulse
response ("RIR") may be captured at a defined location in a
listening room. As an example, an omni-directional test microphone
may be positioned near a loudspeaker, e.g., at a distance of
approximately 0.5-1.5 meters. In general, an excitation signal,
which may be a signal produced by a logarithmic sine sweep, is fed
to the device under test ("DUT"), in this case, the loudspeaker,
and the response of the DUT is captured and compared with the
original signal, as shown in FIG. 1.
[0044] In step 306, the sequence (i.e., the impulse response) is
multiplied by a rectangular or other time window, thus setting
samples above a defined value to t.sub.1 zero (where t.sub.1 is
typically 2-4 milliseconds ("ms") or 100-200 samples at a sample
rate of 48 kHz). This "windowing" suppresses unwanted reflections
from boundaries that are not considered near-field. Next, in step
308, the magnitude spectrum F(i), with i=1, . . . , N/2, is
generated using an N-point FFT, where, for example, N=8192. In step
310, the magnitude spectrum generated in step 308 is smoothed with
a smoothing factor sm.sub.1, resulting in Fs(i)=mean {F(i/sm.sub.1)
. . . F(i*sm.sub.1)}. Typically, the smoothing factor sm.sub.1 may
be equal to approximately 1.05-1.2.
[0045] Proceeding to step 312, the log-magnitude spectrum As of the
inverse system (which is the EQ-filter) is determined by As=-20*log
10(Fs). Next, in step 314, the peaks of As are smoothed with
smoothing factor sm.sub.2, which generally is larger than sm.sub.1
(e.g., sm.sub.2 is typically equal to 1.2-1.6), resulting in Asp
(see plot 610, FIG. 6). This "smoothing of peaks" is illustrated in
FIG. 6. It ensures that the frequency-dependent filter gain does
not exceed values of the average response, while fine details are
preserved below that average response.
[0046] In step 316, the EQ filter is scaled such that its gain is 0
dB at its operating frequency fg (for example, fg=1 kHz; see point
708, FIG. 7). Below fg, the filter response is replaced by the
constant 0 dB. Next, in step 318, the filter response is limited to
its value at a frequency fgu (typically 10-15 kHz), ensuring that
there is no excessive gain to, for example, equalize a tweeter with
a natural roll-off in case the microphone is not positioned exactly
at the main axis. In step 320, filter values above a defined gain
value are set to that defined gain value, in effect, further
limiting the maximum gain of the response and clipping the peaks of
the response.
[0047] In step 322, an EQ filter impulse response is determined
from the scaled, limited, and clipped EQ filter spectrum generated
in steps 316, 318, and 320, assuming minimum-phase. It is
appreciated by those skilled in the art that the EQ filter impulse
response generated in step 322 may be generated using several
techniques, including the Hilbert transform. In step 324, a
rectangular time window is multiplied with the resulting impulse
response according to the desired filter length of, e.g., 64
samples (see point 808, FIG. 8).
[0048] In optional step 326, an equivalent IIR filter impulse
response of low order (typically 2-8) may be generated using a
known method, such as the iterative Steiglitz-McBride method that
approximates the original FIR impulse response in the time domain
by the impulse response of an IIR system (see plot 908, FIG. 9).
(For example, the macro "stmbc," which is part of the MATLAB
package, may be used). The process 300 then ends in step 330.
[0049] A graphical representation 400 of an example of a plot 406
of amplitude 402 (in dBs) versus time 404 (in samples) of a room
impulse response ("RIR") is shown in FIG. 4. The RIR impulse
response, which is captured in step 306, FIG. 3, is multiplied by a
time window 408 for samples above a defined value t.sub.1 such that
these samples are set to zero (see step 308, FIG. 3). Typically,
t.sub.1 may be equal to 2-4 ms or 100-200 samples at a sample rate
of 48 kHz (in FIG. 4, t.sub.1 is equal to approximately 110
samples). This "windowing" suppresses unwanted reflections from
boundaries that are not considered near-field.
[0050] Tuning to FIG. 5, a graphical representation 500 of an
example of plots 506 and 508 of magnitude 502 (in dBs) versus
frequency 504 (in Hz) for the spectrum of the RIR 406 in FIG. 4 is
shown, Plot 506 is the magnitude spectrum F(i), with i=1, . . . ,
N/2, generated using an N-point FFT, where N=8192. Plot 508 is the
magnitude spectrum of plot 506 smoothed with a smoothing factor
sm.sub.1, resulting in Fs(i)=mean {F(i/sm.sub.1) . . .
F(i*sm.sub.1))}. Typically, the smoothing factor sm.sub.1 may be
equal to approximately 1.05-1.2.
[0051] FIG. 6 shows a graphical representation 600 of an example of
plots 606, 608, and 610 of magnitude 602 (in dBs) versus frequency
604 (in Hz) of a frequency response of an ideal EQ filter, a
smoothed version of that frequency response, and the smoothed
version with those parts of the frequency response of an ideal EQ
filter that lie above the smoothed version of the frequency
spectrum cut from the plot, respectively. Plot 606 is a plot of the
log-magnitude spectrum of the inverse system (which is the
EQ-filter) As=-20*log 10(Fs), Plot 608 is a plot of the As of Plot
606 that has been smoothed with smoothing factor sm.sub.2, which
generally is larger than sm.sub.1 (e.g., sm.sub.2 is typically
equal to 1.2-1.6). Cutting that portion of plot 606 that lies above
plot 608 results in plot 610, denoted as Asp. This "smoothing of
peaks" ensures that the frequency-dependent filter gain does not
exceed values of the average response, while fine details are
preserved below that average response.
[0052] In FIG. 7, a graphical representation 700 of an example of a
plot 706 of magnitude 702 (in dBs) versus frequency 704 (in Hz) of
an EQ filter frequency response is shown. The EQ filter generating
the response illustrated by plot 706 has been scaled such that its
gain is 0 dB at its operating frequency fg (at point 708, where fg
is equal to 1 kHz). Below fg, the filter response is replaced by
the constant 0 dB. Above a frequency fgu (at point 710, where fgu
is typically equal to approximately 10-15 kHz), the filter response
is limited to its value at fgu, ensuring that there is no excessive
gain to, for example, equalize a tweeter with a natural roll-off in
case the microphone is not positioned exactly at the main axis. The
maximum gain may be further limited by setting filter values above
a defined gain value to that value (i.e., clipping).
[0053] FIG. 8 shows a graphical representation 800 of an example of
a plot 806 of magnitude 802 (in dBs) versus time 804 (in samples)
of an EQ filter impulse response that is generated from the scaled,
limited, and clipped EQ filter frequency response shown by plot 706
of FIG. 7, assuming minimum-phase. It is appreciated by those
skilled in the art that the EQ filter impulse response depicted by
plot 806 may be generated using several techniques, including the
Hilbert transform. The result of the transform may be time limited
to the desired filter length by applying a rectangular window,
which in FIG. 8 is the length of 64, denoted by point 808.
[0054] In FIG. 9, a graphical representation 900 of an example of
plots 706, FIG. 7, and 908 of magnitude 902 (in dBs) versus
frequency 904 (in Hz) is shown. Plot 706, FIG. 7, depicts the EQ
filter frequency response that has been scaled to frequency fg,
limited above a frequency fgu, and clipped at a maximum gain.
Alternatively, an equivalent IIR filter impulse response of low
order (typically 2-8) may be generated using a known method, such
as the iterative Steiglitz-McBride method that approximates the
original FIR impulse response in the time domain by the impulse
response of an IIR system. (For example, the macro "stmbc," which
is part of the MATLAB package, may be used). An example of an
equivalent IIR filter frequency response is shown by plot 908.
[0055] FIG. 10 shows a graphical representation 1000 of all example
of plots 1006, 1008, and 1010 of magnitude 602 (in dBs) versus
frequency 604 (in Hz) that illustrate the effect of a near-field EQ
on a loudspeaker in a small room. Plot 1008 is a plot of the
log-magnitude frequency response of the loudspeaker obtained in the
near field. Plot 1006 is a plot of the log-magnitude frequency
response of the EQ filter frequency response generated as shown in
FIG. 7 that is applied to the frequency response depicted by plot
1008, with the result being a frequency response depicted by plot
1010. From plot 1010, it is apparent that the measured frequency
response is corrected within the band of interest, i.e., above 1
kHz, where the frequency response is flatter, while less audible,
strongly position-dependent fine details or interference notches
are left unaltered.
[0056] Turning to FIG. 11, a flow chart illustrating another
example of a room equalization method is shown, where the method is
designed for a far-field, low-frequency EQ. The process 1100 may be
a subset of the process 300 shown in FIG. 3, with the following
exceptions. The process starts in step 1102. Next, in step 1104,
the captured frequency response may be multiplied by a "target
function" in order to obtain the ideal EQ filter response.
Typically this may be a bandpass filter with a passband of 20-80 Hz
(e.g., a 4.sup.th order Butterworth characteristic). More complex
target functions may be utilized, particularly in automotive
applications.
[0057] Step 306, FIG. 3, where the sequence (impulse response) is
multiplied by a rectangular or other time window, is not included
in process 1100 because correction of the complete room impulse
response ("RIR") is possible and also desirable at low frequencies.
Smoothing of peaks, however, applies similarly as in the
near-field, HF-EQ process and this takes place in step 1106. In
step 1108, the resulting FIR filter may be scaled to an average
loudness level, and directly implemented at a lower sample rate
(typically 375 Hz, which corresponds to a decimation ratio of 64 at
a frequency of 48 kHz) using decimation and interpolation filters,
as shown by decimation filter 208 and interpolation filter 214,
FIG. 2. FIG. 12 shows a graphical representation 1200 of an example
of a plot 1206 of magnitude 1202 (in dBs) versus frequency 1204 (in
Hz) of a typical Bass EQ filter frequency response.
[0058] A mid-frequency ("MF") EQ operates in the frequency range
of, for example, 100 Hz-1 kHz. Room impulse responses may be
captured by a microphone that is located at the desired listening
location. In FIG. 13, a flow chart illustrating an example of
another room equalization method is shown, where this method is
designed for a far-field, mid-frequency EQ. The process 1300 starts
in step 1302 and in step 1304, a room impulse response ("RIR") may
be determined at a listening location, Steps 1304, 1306, 1308,
1310, and 1312 are similar to the corresponding steps of FIG. 3;
however, the parameters are chosen differently.
[0059] In step 1306, the sequence (i.e., the impulse response) is
multiplied by a rectangular or other time window, thus setting
samples above a defined value t.sub.2 to zero. This time windowing
now has a larger impact, because major parts of the measured
impulse response are cut off (see FIG. 14). As a result, only the
source (i.e., the loudspeaker) and its direct adjacent surfaces are
included, thus focusing on source, not room, correction. This leads
to increased robustness with respect to microphone placement, and
thus optimum correction over the entire listening area, not just a
single point.
[0060] Next, in step 1308, the magnitude spectrum F(i), with i=1, .
. . , N/2, is generated using an N-point FFT, where, for example,
N=8192, In step 1310, the magnitude spectrum determined in step
1308 is smoothed with a smoothing factor sm.sub.3, resulting in
Fs(i)=mean {F(i/sm.sub.3) . . . F(i*sm.sub.3)}. Typically, the
smoothing factor sm.sub.3 used in the far-field, MF EQ, is much
larger than the smoothing factor used in the HF EQ (typically,
sm.sub.3=1.4-2.0), so that only the overall trend will be
considered, not fine details. Also, the MF EQ does not apply
separate smoothing of peaks and dips, as shown in step 314, FIG.
3.
[0061] In step 1312, the logarithmic magnitude spectrum is
determined and normalized to a prescribed maximum gain. In step
1314, the EQ filter frequency response may be determined by
negating the log-magnitude spectrum of step 1312 and adding a
high-pass target function (typically, 80-200 Hz), and in step 1316,
the EQ filter frequency response is set to zero dB above its
operating range. The process 1300 then ends in step 1320.
[0062] FIG. 14 shows a graphical representation 1400 of an example
of a plot 1406 of amplitude 1402 (in dBs) versus time 1404 (in
samples) of the RIR generated in step 1304 of FIG. 1304. The RIR is
multiplied by a time window 1408 for samples above a defined value
t.sub.2 such that these samples are set to zero. Typically, t.sub.2
may be equal to 16 . . . 32 millisecs ("ms") or 100-200 samples at
a sample rate of 8 kHz (in FIG. 4, t.sub.2 is equal to
approximately 130 samples). As noted above when discussing FIG. 13,
this "windowing" cuts off major parts of the RIR.
[0063] Turning to FIG. 15, a graphical representation 1500 of an
example of spectral plots 1506 and 1508 of amplitude 1502 (in dBs)
versus frequency 504 (in Hz) for the RIR 1406 of FIG. 14 is shown.
Plot 1506 is the amplitude spectrum F(i), with i=1, . . . , N/2,
computed using an N-point FFT, where N=8192. Plot 1508 is the
amplitude spectrum of plot 1506 smoothed with a smoothing factor
sm.sub.3, resulting in Fs(i)=mean {F(i/sm.sub.3) . . .
F(i*sm.sub.3)}. As noted above when discussing FIG. 13, the larger
smoothing coefficient sm.sub.3 generates a plot 1508 that takes
into account only the overall trend, not fine details.
[0064] FIG. 16 shows a graphical representation 1600 of an example
of plots 1606 and 1608 of amplitude 1602 (in dBs) versus frequency
1604 (in Hz), where plot 1606 is a plot of the smoothed
log-magnitude spectrum of the measured response and plot 1608 is a
plot of the EQ filter impulse response obtained using a target high
pass function. Turning to FIG. 17, a graphical representation 1700
of an example of plots 1706 and 1708 of amplitude 1702 (in dBs)
versus frequency 1704 (in Hz) is shown, Plots 1706 and 1708 are the
frequency responses of low- and mid-frequency EQ filters,
respectively. FIG. 18 shows a graphical representation 1800 of all
example of plots 1806, 1808, and 1810 of amplitude 1802 (in dBs)
versus frequency 1804 (in Hz), where plot 1806 is a plot of the
inverse system, plot 1808 is a plot of the log-magnitude spectrum
that has been smoothed with a smoothing factor, and plot 1810 is
the sum of 1806 and 1808, shifted downwards for better visibility,
showing the result after EQ.
[0065] In automotive applications, it is no longer necessary, or
desirable, to distinguish between near- and far-field responses.
More complex target functions, such as that shown in FIG. 19, may
be utilized in order to predict average responses at the automobile
seats that include direct and reflected sound fields. FIG. 19 shows
a graphical representation 1900 of an example of a plot 1906 of
magnitude 702 (in dBs) versus frequency 704 (in Hz) of an EQ filter
frequency response generated using another example of a target
function. The equalization may be performed as described, using
different smoothing factors in different frequency bands. Input
data may be obtained by spatial averaging between different
locations around the listener's head, and between the seats. Also,
weighting factors may be applied to emphasize equalization quality
at a particular seat, while compromising performance at other
seats.
[0066] In order to save processing costs and minimize complexity,
equalization may be performed throughout the whole frequency band
at once. However, the resulting filter impulse response may be
split into several bands, as shown in FIG. 20. In FIG. 20, a
graphical representation 2000 of an example of plots 2006, 2008,
and 2010 of magnitude 2002 (in dBs) versus frequency 2004 (in Hz)
of EQ filter impulse responses is shown, Plots 2006, 2008, and 2010
depict the frequency spectra for the low, medium, and high
frequency bands, respectively. It is then easier to approximate the
individual, band-limited responses separately by low-order IIR
filters using, for example, the Steiglitz-McBride method as
described earlier. The resulting individual EQ-sections may then be
connected in series,
[0067] Persons skilled in the art will understand and appreciate,
that one or more processes, sub-processes, or process steps
described in connection with FIGS. 3, 11, and 13 may be performed
by hardware and/or software. Additionally, the WBES described above
may be implemented completely in software that would be executed
within a processor or plurality of processors in a networked
environment. Examples of a processor include but are not limited to
microprocessor, general purpose processor, combination of
processors, DSP, any logic or decision processing unit regardless
of method of operation, instructions
execution/system/apparatus/device and/or ASIC. If the process is
performed by software, the software may reside in software memory
(not shown) in the device used to execute the software. The
software in software memory may include an ordered listing of
executable instructions for implementing logical functions (i.e.,
"logic" that may be implemented either in digital form such as
digital circuitry or source code or optical circuitry or chemical
or biochemical in analog form such as analog circuitry or an analog
source such an analog electrical, sound or video signal), and may
selectively be embodied in any signal-bearing (such as a
machine-readable and/or computer-readable) medium for use by or in
connection with an instruction execution system, apparatus, or
device, such as a computer-based system, processor-containing
system, or other system that may selectively fetch the instructions
from the instruction execution system, apparatus, or device and
execute the instructions. In the context of this document, a
"machine-readable medium," "computer-readable medium," and/or
"signal-bearing medium" (herein known as a "signal-bearing medium")
is any means that may contain, store, communicate, propagate, or
transport the program for use by or in connection with the
instruction execution system, apparatus, or device. The
signal-bearing medium may selectively be, for example but not
limited to, an electronic, magnetic, optical, electromagnetic,
infrared, or semiconductor system, apparatus, device, air, water,
or propagation medium. More specific examples, but nonetheless a
non-exhaustive list, of computer-readable media would include the
following: an electrical connection (electronic) having one or more
wires; a portable computer diskette (magnetic); a RAM (electronic);
a read-only memory "ROM" (electronic); an erasable programmable
read-only memory (EPROM or Flash memory) (electronic); an optical
fiber (optical); and a portable compact disc read-only memory
"CDROM" (optical). Note that the computer-readable medium may even
be paper or another suitable medium upon which the program is
printed, as the program can be electronically captured, via, for
instance, optical scanning of the paper or other medium, then
compiled, interpreted or otherwise processed in a suitable manner
if necessary, and then stored in a computer memory. Additionally,
it is appreciated by those skilled in the art that a signal-bearing
medium may include carrier wave signals on propagated signals in
telecommunication and/or network distributed systems. These
propagated signals may be computer (i.e., machine) data signals
embodied in the carrier wave signal. The computer/machine data
signals may include data or software that is transported or
interacts with the carrier wave signal.
[0068] While the foregoing descriptions refer to the use of a wide
band equalization system in smaller enclosed spaces, such as a home
theater or automobile, the subject matter is not limited to such
use. Any electronic system or component that measures and processes
signals produced in an audio or sound system that could benefit
from the functionality provided by the components described above
may be implemented as the elements of the invention.
[0069] Moreover, it will be understood that the foregoing
description of numerous implementations has been presented for
purposes of illustration and description. It is not exhaustive and
does not limit the claimed inventions to the precise forms
disclosed. Modifications and variations are possible in light of
the above description or may be acquired from practicing the
invention. The claims and their equivalents define the scope of the
invention.
* * * * *